Computer Upgrade - Normal Singing Voices Transformed
A new technology can convert everyday human voices into beautiful ones. Will this make karaoke more or less fun? And was Bart Simpson behind this?

In an episode of the Simpsons first aired in February, 2001, Bart and his friends are recruited into a new "boy band" by an unscrupulous promoter. The fact that the kids couldn't actually sing didn't matter -- he simply ran their voices through a "voice enhancer," and the Party Posse could bust their moves without busting anyone's eardrums.

Life imitates art. Referring to research he recently presented at the 145th meeting of Acoustical Society of America, Prof. Mark Smith of Purdue University said, "Our [project's] ultimate goal is to have a computer system that will transform a poor singing voice into a great singing voice."

Smith, a former faculty member at the Georgia Institute of Technology, now heads Purdue's School of Electrical and Computer Engineering. Working with Georgia Tech graduate student Matthew Lee, he has created computer models for voice analysis and synthesis that break the human singing voice into components that can then be modified to produce a more professional-sounding rendition of the original.

The system breaks down the original voice into its sine wave components, which are then tweaked in the right direction to fix problems in pitch, duration and vibrato (the natural pitch warble that makes a pro sound good). The voice is then reconstructed using those enhancements. And it all is done "real time," or at least quickly enough to stay in time.

The system is still a long way from appearing at your local karaoke bar. "While we have had success in improving the quality of the singing voice samples in our database, we have a way to go before we are able to handle all types of voices reliably," Smith said.

The challenges ahead include dealing with the wide variety of singing styles & voice types, tackling gender differences (male voices are harder to fix, it turns out, because of their lower pitch), making the result sound natural and not choppy, and having the resulting "good voice" resemble the original as closely as possible.

Other possible applications of Smith & Lee's model include synthesizing musical instruments, improving the quality of text-to-speech programs, and better programs for the hearing-impaired.

Read Purdue's press release about the research, and hear MP3 files of voices before and after the mathematical treatment

Explore the ASA site, including "some interesting sounds" (such as whale cries, an exploding meteor, and the "Blue Danube")

Learn more at ASA’s service "Acoustics.org," which contains many sound resources, including demos, software, presentations, bibliographies

Revisit the Feb. 2001 Simpsons episode "New Kids on the Blech," featuring the "Party Posse." Yvan eht Nioj, baby!